Benchmark
Tau2 Telecom
communication
reasoning
text
τ²-Bench telecom domain evaluates conversational agents in a dual-control environment modeled as a Dec-POMDP, where both agent and user use tools in shared telecommunications troubleshooting scenarios that test coordination and communication capabilities.
语言EN
满分1
参评模型9
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | GPT-5 | OpenAI | 96.7 | 来源 ↗ |
| 2 | Claude Haiku 4.5 | Anthropic | 83.0 | 来源 ↗ |
| 3 | Kimi K2 Instruct | Moonshot AI | 65.8 | 来源 ↗ |
| 4 | Kimi K2-Instruct-0905 | Moonshot AI | 65.8 | 来源 ↗ |
| 5 | o3 | OpenAI | 58.2 | 来源 ↗ |
| 6 | Qwen3-235B-A22B-Thinking-2507 | Alibaba Cloud / Qwen Team | 45.6 | 来源 ↗ |
| 7 | Qwen3-Next-80B-A3B-Thinking | Alibaba Cloud / Qwen Team | 43.9 | 来源 ↗ |
| 8 | GPT-4o | OpenAI | 23.5 | 来源 ↗ |
| 9 | Qwen3-Next-80B-A3B-Instruct | Alibaba Cloud / Qwen Team | 13.2 | 来源 ↗ |